A park multi-energy grid multi-energy mutual aid method for grid interaction

Abstract

The application discloses a kind of park multi-energy network multi-energy mutual aid method for grid interaction, belong to the field of power grid dispatching, comprising the following steps: S1, construct time sequence characteristic library;S2, adopt Latin hypercube sampling-Jordan decomposition-K means clustering time sequence coupling scene generation and double-layer reduction method, generate multi-energy load typical scene set;S3, construct the double-layer-distributed hybrid optimization model of park multi-energy network driven by grid interaction;S4, decouple multi-energy flow subsystem using dual decomposition method, obtain subsystem optimal solution;S5, check subsystem optimal solution, output park multi-energy network optimal mutual aid scheme.The above-mentioned park multi-energy network multi-energy mutual aid method for grid interaction not only realizes multi-energy flow subsystem decoupling collaborative modeling, improves intelligent optimization algorithm optimization precision and convergence speed, but also considers park multi-energy network economic operation, power grid safety and stability, peak-valley suppression and low-carbon quota control multiple demand.

Description

Technical Field

[0001] This invention relates to the field of power grid dispatching technology, and in particular to a multi-energy mutual assistance method for multi-energy networks in industrial parks oriented towards power grid interaction. Background Technology

[0002] With the large-scale popularization of integrated energy parks, various types of energy equipment, such as photovoltaics, wind power, gas turbines, energy storage, electrolytic hydrogen production, and combined cooling and heating, are densely connected to the park's multi-energy grid, forming a deep coupling operation mode of electricity, heat, cooling, and hydrogen. The park's multi-energy grid is no longer limited to internal source-load power matching and economic dispatch, but also needs to adapt to the grid connection requirements of the external large power grid, meet the requirements of grid peak-valley difference suppression, system equivalent inertia support, and frequency change rate fluctuation limit. At the same time, it is constrained by carbon emission quota control policies, exhibiting characteristics of strong multi-energy coupling, high optimization variable dimension, and multiple safety and low-carbon constraints. This places stringent requirements on the modeling architecture, intelligent optimization algorithm, multi-constraint adaptation capability, and compliance verification mechanism of the park's multi-energy grid distributed optimization dispatch.

[0003] Currently, the field of multi-energy grid optimization and scheduling in industrial parks generally adopts centralized traditional evolutionary algorithms and conventional decomposition and coordination optimization techniques. The algorithm initialization is mostly done by pure random generation. The population architecture is mostly a single overall population. At the modeling level, a unified and general power balance constraint is mostly used. This results in low solution accuracy, slow convergence speed, and easy getting trapped in local optima under high-dimensional and complex scenarios. The power interaction boundary is vaguely defined, and it cannot be synchronously adapted to the actual operating conditions of grid safety connection, peak-valley difference regulation, and carbon emission quota management. Summary of the Invention

[0004] The purpose of this invention is to provide a method for multi-energy mutual assistance in a multi-energy network in a park, oriented towards power grid interaction, to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, this invention provides a multi-energy network mutual assistance method for campuses oriented towards grid interaction, comprising the following steps: S1, constructing a time-series feature library by integrating multi-dimensional data from the energy side, grid side, equipment side, load side, and interaction side; S2, based on the time-series feature library output by S1, generating a typical multi-energy load scenario set and corresponding probabilities for grid interaction using a Latin hypercube sampling-Cholsky decomposition-K-means clustering time-series coupled scenario generation and two-layer reduction method; S3, based on the typical scenario set output by S2, constructing a grid interaction-driven two-layer-distributed hybrid optimization model for campus multi-energy networks; S4, based on the hybrid optimization model constructed by S3 and the TG-OMPE-QPSO hybrid algorithm, decoupling the multi-energy flow subsystems using the dual decomposition method to obtain the optimal solution for the subsystems; S5, verifying the optimal solution for the subsystems output by S4, and outputting the optimal mutual assistance scheme for the campus multi-energy network.

[0006] Therefore, the present invention employs the above-mentioned method for multi-energy grid mutual assistance in a campus oriented towards power grid interaction, which has the following beneficial effects: 1. Layered decoupling of modeling architecture significantly improves logical rigor: The multi-energy network of the park is decoupled into eight independent units: wind power system, photovoltaic subsystem, park power subsystem, thermal subsystem, cold subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem. It clearly distinguishes the two types of balance constraints: global grid interaction power and internal grid power. It independently defines the cross-subsystem coupling constraints of multi-energy flows such as heat, cold, hydrogen, and energy storage, realizes the boundary separation between internal grid security control and external grid power purchase and sales interaction, and clarifies the power and energy interaction relationship of multi-energy flows. The optimized model fits the actual park operation architecture. 2. Orthogonal array initialization for more uniform solution space coverage: The decision variables are uniformly discretized by using a fixed orthogonal discretization level and a custom orthogonal array type, replacing the traditional random population initialization method. This allows the initial population to uniformly traverse and optimize the solution space, avoiding blind initialization and improving the algorithm's early optimization traversal ability and solution stability. 3. Enhanced Algorithm Optimization Performance with Diverse Population Ratios and Gaussian Perturbation: By setting a fixed ratio structure for the basic population and elite population, and combining it with a Gaussian random walk perturbation mechanism, the perturbation standard deviation is adaptively set according to the feasible region of each decision variable. Adaptive perturbation is achieved through perturbation probability, boundary coefficient, and weight adjustment coefficient. This allows the algorithm to escape local optima when it stops iterating, enriching population diversity and significantly improving global search capabilities and the quality of optimal solutions. 4. Multi-constraint collaborative verification to adapt to actual grid-connected operation requirements: Construct a five-in-one collaborative constraint verification system that integrates grid peak-valley difference suppression rate, system equivalent inertia, frequency change rate, global power deviation, and carbon quota. This breaks through the limitations of traditional single economic constraints and simultaneously takes into account multiple objectives such as grid peak-valley regulation, power system safety and stability, power stability and low-carbon quota management, making the output dispatch scheme more practical for engineering. 5. Closed-loop iterative verification mechanism, balancing compliance and computational efficiency: Setting constraints to backtrack and restart orthogonal multi-population evolution iteration when unqualified, and combining it with stagnation iteration threshold termination rules, forming a closed-loop logic of "optimization-verification-unqualified re-optimization-compliant output", which strictly ensures that all final candidate solutions meet multi-dimensional operation constraints, while avoiding infinite algorithm iteration, thus balancing solution compliance and computational convergence efficiency.

[0007] In summary, the overall technical solution of this invention constructs a two-layer distributed optimization scheduling system adapted to multi-energy networks in industrial parks, combining orthogonal multi-population evolution with dual decomposition. By decomposing the multi-energy network into eight types of distributed subsystems and accurately distinguishing between the global power grid interaction energy coupling constraints and the internal power grid energy coupling constraints of the power subsystems, the system completes the uniform discretization of decision variables and population initialization based on fixed orthogonal discretization levels and orthogonal arrays. Gaussian random walk perturbation is introduced to achieve adaptive global search of the algorithm. At the same time, a multi-dimensional collaborative constraint verification system including power grid peak-valley difference, equivalent inertia, frequency change rate, global power deviation, and carbon quota is built. A closed-loop mechanism for backtracking and iterative re-optimization of unqualified solutions is set up. This not only realizes the decoupling and collaborative modeling of multi-energy subsystems and improves the optimization accuracy and convergence speed of intelligent optimization algorithms, but also takes into account the multiple needs of economic operation of multi-energy networks in industrial parks, power grid safety and stability, peak-valley suppression, and low-carbon quota management. It effectively makes up for the shortcomings of traditional multi-energy network optimization scheduling in industrial parks in terms of modeling architecture, algorithm performance, constraint adaptation, and verification logic.

[0008] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0009] Figure 1 This is a flowchart of a multi-energy mutual assistance method for a multi-energy network in a park, oriented towards power grid interaction, as described in this invention. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0011] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0012] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0013] like Figure 1As shown, a multi-energy network mutual assistance method for campuses oriented towards grid interaction includes the following steps: S1, constructing a time-series feature library by integrating multi-dimensional data from the energy side, grid side, equipment side, load side, and interaction side; S2, based on the time-series feature library output by S1, generating a typical multi-energy load scenario set and corresponding probabilities for grid interaction using a Latin hypercube sampling-Cholsky decomposition-K-means clustering time-series coupled scenario generation and two-layer reduction method; S3, based on the typical scenario set output by S2, constructing a grid interaction-driven two-layer-distributed hybrid optimization model for campus multi-energy networks; S4, based on the hybrid optimization model constructed by S3 and the TG-OMPE-QPSO hybrid algorithm, decoupling the multi-energy flow subsystems using the dual decomposition method to obtain the optimal solution for the subsystems; S5, verifying the optimal solution for the subsystems output by S4, and outputting the optimal mutual assistance scheme for the campus multi-energy network.

[0014] Step S1 specifically includes the following steps: S11. Simultaneously collect multi-dimensional data from the energy side, grid side, equipment side, load side, and interaction side; among which, the energy side data includes photovoltaic output. Wind power output Gas turbine output Hydrogen production capacity ; Grid-side data includes time-of-use pricing. Carbon trading prices Grid peak shaving depth Minimum inertia threshold ; Equipment-side data includes parameters for photovoltaic modules, wind turbine generators, gas turbines, combined heat and power (CHP) systems, electrochemical energy storage systems, thermal energy equipment, cooling energy equipment, and water electrolysis for hydrogen production and hydrogen coupling equipment. Photovoltaic module parameters include rated installed capacity. Photovoltaic module conversion efficiency Photovoltaic power temperature coefficient Irradiation intensity under standard test conditions With temperature upper and lower limits of photovoltaic power output constraints , Photovoltaic module floor area Operation and maintenance loss coefficient Photovoltaic curtailment power Photovoltaic operation and maintenance costs and the cost of solar curtailment penalties Wind turbine generator parameters include rated installed capacity. Fan cut-in wind speed Rated wind speed Cut-off wind speed Upper and lower limits of wind turbine output constraints , Wind energy utilization coefficient, wind turbine transmission efficiency The floor area of ​​the fan unit Wind turbine operation and maintenance cost coefficient Gas turbine parameters include rated power generation. Power generation efficiency upper and lower limits of gas turbine output constraints , Carbon emission factors throughout the entire life cycle Gas turbine operation and maintenance costs Cogeneration characteristic parameters include heat-to-power ratio. Heating efficiency Gas turbine floor space ; Electrochemical energy storage system parameters include rated capacity Rated charge and discharge power Charging efficiency Discharge efficiency Discharge power Charging power Self-discharge rate Upper and lower limits of charge state constraints , Depth of energy storage Cycle life mapping parameters, energy storage degradation cost coefficient Energy storage charging and discharging loss cost Energy storage operation and maintenance costs Energy storage unit floor area Charging power upper and lower limits , upper and lower limits of discharge power , ; Thermal energy equipment parameters include the rated heating power of the ground source heat pump. Coefficient of performance (COP) Output upper and lower limits and minimum operating load ratio, heating efficiency of gas-fired boilers Heating power of gas boiler Operation and maintenance costs of thermal energy equipment Thermal energy equipment floor area ; Parameters of cooling equipment include the rated cooling capacity of the ground source heat pump. Coefficient of performance (COD) Output upper and lower limits and minimum operating load ratio, refrigeration efficiency of electric refrigeration unit Electric cooling power Operation and maintenance costs of cold energy equipment Refrigeration equipment floor space ; Parameters of water electrolysis for hydrogen production and hydrogen energy coupling equipment include the rated hydrogen production power of the electrolyzer. Electro-hydrogen conversion efficiency Upper and lower limits of hydrogen production capacity constraints , Hydrogen energy storage rated pressure, capacity constraints and hydrogen energy utilization efficiency, and hydrogen production equipment operation and maintenance costs. and carbon emission factors Hydrogen storage and release power Hydrogen storage and charging power ; Load-side data includes electrical load. Heat load Cooling load Hydrogen load Time series data; Interactive data includes the power purchased by the park's power grid. Power sales volume of the park's power grid Total power exchanged between the power grid Grid interaction response commands, upper and lower limits of grid tie line power , ; S12. Perform data cleaning, neighborhood mean interpolation for missing values, and normalization on the original multidimensional data to obtain standardized multidimensional data. S13. Fit the probability distribution to the standardized multidimensional data to obtain the probability distribution model; S14. The Kendall rank correlation coefficient is used to calculate the time-series correlation of the source-load-storage multi-energy units of the multi-energy grid in the park, and the source-load-storage time-series correlation matrix is ​​obtained. S15. Integrate the probability distribution model and the source-load-storage time-series correlation matrix to construct a time-series feature library for the park's multi-energy network.

[0015] Step S2 specifically includes the following steps: S21. Based on the time-series feature library output by S1, extract the park's features. Heat load Cooling load Hydrogen load Photovoltaic power output Wind power output Gas turbine output Power, hydrogen production capacity Total power grid interaction Nine sampling variables are used as sampling dimensions to obtain the sampling variable dimension set. The distribution function of each sampling variable was determined, the source-load-storage time-series correlation matrix was extracted, and a standardized time-series correlation coefficient matrix was constructed. ; S22. Latin hypercube sampling is used to achieve multivariate stratified unbiased sampling and generate independent initial scenarios. S221. Based on the sampling variable dimension set and empirical distribution function set output by S21, the probability distribution interval of each sampling variable is determined. Divided into equal parts There are three non-overlapping sub-intervals, each with a width of [value missing]. ; S222. Independently and randomly select uniformly distributed random numbers within each sub-interval, and calculate the sampling probability value of the sub-interval: ; In the formula, For the first The sampling probability value of each sub-interval; The numbers are uniformly distributed random numbers, and ; S223. Generate the first... through inverse transformation of empirical distribution. In the initial sampling scenario, the first The value of each variable : ; In the formula, For the first The sampling variable of the first One initial sampled scenario value; For the first The inverse function of the empirical distribution of each variable; S224. Arrange the initial scene values ​​of all variables according to the scene number and assemble them into a multi-energy load independent initial sampling scene matrix. , The total dimension of the source load random variables participating in the scenario sampling of the multi-energy network in the park. The total number of original, unreduced scenes generated by Latin hypercube sampling; S225. Determine the independent initial sampling scenario matrix. If the Euclidean distance between any two scenes is greater than a preset threshold, then the scene independence is deemed satisfactory, and the matrix of the independent initial sampled scenes is output. Otherwise, discard the sample and return to S221 to re-execute the sampling process; S23. Choleski decomposition to correct temporal coupling correlation: based on the independent initial sampling scene matrix output in step S22. The standardized time series correlation coefficient matrix output by S21 The temporal coupling relationship of the source load storage variables is restored and corrected to obtain the initial scene set of temporal coupling; S24. Initial Scenario Standardization and Power Grid Interaction Tag Matching: Based on the time-coupled initial scenario set output by S23, the scenario data is processed by maximum and minimum normalization. Then, combined with the power grid interaction time series parameters, the interaction constraint requirements are labeled for each scenario, and a standardized initial scenario set with power grid interaction tags is output. S25, First-level scene reduction: Based on the standardized initial scene set output by S24, the K-means clustering algorithm is used to reduce the dimensionality of the scenes and output the scene set with K-means cluster centers; S26, Second-layer scene reduction: Based on the K-means cluster center scene set output by S25, redundant scenes are eliminated using the probability distance criterion until the number of scenes is reduced to the preset number of typical scenes. The reduced candidate typical scene set and the initial probability value of the scene are output. S27. Scene probability redistribution and normalization: Based on the candidate typical scene set output by S26, the probability of the eliminated scene is redistributed and superimposed on the most similar retained scene. The probability of all retained scenes is normalized and corrected to ensure that the sum of the probabilities of all scenes is 1. S28. Scene Validity Verification and Error Assessment: Based on the typical scene set output in S27, combined with the original multi-dimensional data output in S1, the fitting accuracy of the scene is evaluated using the average fitting error as an indicator, and the temporal correlation retention of the scene is calculated. : ; In the formula, The temporal correlation coefficients for a typical set of scenarios; If the average fitting error of the scene And time-series correlation retention If the scenario is valid, the final set of typical multi-energy load scenarios, the corresponding scenario probabilities, and the validity verification report will be output.

[0016] Step S2 addresses the strong coupling of time-series fluctuations in the multi-energy flow (electricity / heat / cold / hydrogen) and the two-sided temporal fluctuations (source / load / storage) of the multi-energy grid in the park. It employs Latin hypercube sampling (LHS) to achieve multivariate hierarchical unbiased sampling and uses Cholesky decomposition to restore the true temporal correlation between source, load, and storage, thus solving the problem of neglecting temporal coupling in traditional scenario generation. Furthermore, a two-layer scenario reduction mechanism combining K-means clustering and probabilistic distance is designed to eliminate redundant scenarios while preserving temporal features and grid interaction adaptability, significantly reducing the computational load of subsequent optimization. At the same time, it ensures full coverage of typical operating conditions such as grid peak and valley and new energy fluctuations, providing high-precision, low-redundancy uncertainty input for the S3 two-layer-distributed optimization model.

[0017] The hybrid optimization model described in step S3 has an upper layer of capacity configuration-grid interaction parameter optimization and a lower layer of multi-energy flow distributed operation optimization. It embeds carbon cost, energy storage lifetime, and minimum inertia collaborative constraints, and clarifies the hierarchical positioning of the upper layer towards long-term planning and decision-making and the lower layer towards short-term real-time operation.

[0018] Step S3 specifically includes the following steps: S31. Construction of the upper-level optimization objective function: Based on the typical scene set and scene probability output by S2, construct the upper-level optimization three objective function; The first objective function is expressed as follows: ; In the formula, This represents the park's net income throughout its entire lifecycle. This represents the total number of typical scenarios output by S2. For the first The probability of a typical scenario; To optimize the total timing length; for Total carbon emissions of the park at any given time; for The cost of equipment maintenance at all times, and ; The expression for the second objective function is as follows: ; in, ; In the formula, This refers to the peak-valley difference in power exchange between power grids. The third objective function is as follows: ; In the formula, This represents the total carbon emissions of the system. Carbon emission factors from purchasing electricity from the power grid; S32. Add upper-level optimization constraints: Equipment capacity upper and lower limits constraints: Power constraints in grid interaction: ; Total investment and land use constraints: ; Constraints on curtailment of solar power: In the formula, For the first Installed capacity of this type of equipment; For the first Lower limit of capacity for this type of equipment; For the first Maximum capacity of this type of device; and These are the upper and lower limits for electricity purchases from the power grid; and These are the upper and lower limits for electricity sales from the power grid; This represents the total investment in equipment within the park. This is the upper limit for investment; The total area is, and ; This is the upper limit for land area; This is the upper limit of the light rejection rate; S33. Decoupling of Lower-Level Distributed Optimization Model and Construction of Subsystem Objectives: Based on the capacity configuration and grid interaction parameters output by the upper-level optimization, the dual decomposition method is used to decouple the global multi-energy flow optimization into eight types of distributed subsystems: wind power system, photovoltaic subsystem, park power subsystem, thermal subsystem, cold subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem. The objective function of the photovoltaic subsystem is as follows: ; In the formula, Operating costs of the photovoltaic subsystem; The objective function of the park's power subsystem is as follows: ; In the formula, The comprehensive objective function value of the park's power subsystem; for The cost of integrated operation, maintenance, and dispatch of the power grid within the industrial park at all times; The internal network power fluctuation penalty weighting coefficient; for Real-time fluctuations in aggregated power output within the park; for Equivalent inertia of the power subsystem in the park at any given moment; The real-time frequency change rate of the park's power subsystem; This represents the upper limit of the frequency change rate of the power subsystem. and These are the large penalty coefficients for exceeding the inertia limit and exceeding the frequency change rate limit, respectively. This is the function for determining whether a function exceeds its limit. The objective functions for the wind electronic system, thermal electronic system, cold electronic system, and hydrogen electronic system are as follows: ; In the formula, For the first The real-time operating cost of each subsystem, the first Each subsystem can be a wind electronic system, a thermal subsystem, a cold subsystem, or a hydrogen subsystem; For the first Cost function of each subsystem; For the first Subsystem Always put in the effort; The objective function of the energy storage subsystem is as follows: ; In the formula, Cost of the energy storage subsystem; The objective function of the power grid interaction subsystem is as follows: ; In the formula, For grid interaction costs; Output the distributed subsystem partitioning results and the single-objective operating cost function; S34. Add lower-level multi-energy flow power balance coupling constraints; Global power grid interaction and energy coupling constraints in the park: ; Thermal energy balance constraints: ; Power coupling constraints within the park's power subsystem: ; Cold energy balance constraints: ; Hydrogen energy balance constraints: ; Energy storage balance constraints: ; In the formula, It contributes to the integration of photovoltaic, wind power, and gas turbine industries in the industrial park; and They are respectively Time and The state of charge of the stored energy at any given moment; To optimize the time step; S35. Based on the interaction relationship of energy storage charging and discharging power in the power balance constraint of S34, construct the energy storage operation-lifetime collaborative constraint. Upper and lower limits of energy storage state of charge constraints: ; Energy storage cycle life constraints: ; In the formula, This represents the maximum number of energy storage cycles. S36. Based on the multi-energy flow power balance coupling constraint in step S34 and the energy storage operation-lifetime synergistic constraint in step S35, embed the grid interaction-low carbon-inertia synergistic constraint. System frequency change rate constraint: ; Carbon quota constraints: ; System equivalent inertia constraint: ; In the formula, The equivalent inertia of the power subsystem; for Carbon emission allowances for the industrial park at any time; S37. Based on the full-dimensional constraint set constructed in S35-S36, the Lagrange relaxation method is used to decouple all variables and construct a global dual function. : ; in, ; ; ; ; ; ; In the formula, This is a set of global decision variables, which includes photovoltaic power output. Wind power output Gas turbine output Hydrogen production capacity Energy storage charging and discharging power, total grid interconnection power , heating and cooling load power; These are the Lagrange multipliers for power balance, thermal balance, park power grid balance, cold energy balance, hydrogen energy balance, and energy storage balance, respectively. The total cost of the wind power system, photovoltaic subsystem, park power subsystem, thermal subsystem, cooling subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem; , , , , , These are the global power supply and demand balance deviation, thermal energy supply and demand balance deviation, power subsystem grid power supply and demand balance deviation, cold energy supply and demand balance deviation, hydrogen energy supply and demand balance deviation, and energy storage system energy balance deviation. This is the native electrical load within the park; For the interface power exchange between the park's power subsystem and the power grid interaction subsystem; Complete the model feasibility verification and output a complete dual-layer-distributed hybrid optimization model for the multi-energy network in the park.

[0019] Step S4 specifically includes the following steps: S41. Dual Decomposition Method for Coupling Constraint Relaxation and Solution Framework Initialization: Based on the campus multi-energy grid two-layer-distributed hybrid optimization model, Lagrange dual function, and full-dimensional constraint set output from S3, the dual decomposition method is used to relax the global grid interaction energy coupling constraints, the campus power subsystem intragrid energy coupling constraints, and the multi-energy flow cross-subsystem coupling constraints of heat, cold, hydrogen, and energy storage. The distributed solution framework is initialized, defining the independent solution boundaries, inter-subsystem communication rules, and global iteration accuracy thresholds for the wind power system, photovoltaic subsystem, thermal subsystem, cold subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem. The Lagrange multiplier vectors are initialized, and the coupling constraint relaxation results, subsystem solution boundary sets, and initial Lagrange multiplier vectors are output. , These are the initial multipliers for electrical energy balance, thermal energy balance, park power grid balance, cold energy balance, hydrogen energy balance, and energy storage balance. S42, TG-OMPE-QPSO Hybrid Algorithm Core Parameter Configuration and Population Initialization: Based on the subsystem boundary solution output from S41, the core operating parameters of the TG-OMPE-QPSO hybrid algorithm are configured; the algorithm population is initialized using a dual mechanism of Tent chaotic mapping followed by orthogonal design, resulting in the Tent-orthogonal hybrid initial population. To ensure that the population covers the feasible area evenly; S43. Encoding of decision variables and feasible region mapping for multi-energy flow subsystems: solving the subsystem boundary based on the output of S41 and the Tent-orthogonal mixed initial population based on the output of S42. The decision variables of the eight subsystems are encoded with real numbers; then, the encoded individuals of the population are mapped to the feasible domains of each subsystem, and matched with the full-dimensional constraint set constructed by S3 to obtain the feasible population of the subsystem; and the decision variable mapping formula is as follows: ; In the formula, For the first Initial population of each subsystem; For the first Each subsystem constraint set; S44, Parallel Optimization of Multiple Subsystems (TG-OMPE-QPSO Iterative Solution): Based on the feasible population of subsystems output by S43, the TG-OMPE-QPSO hybrid algorithm is used to perform parallel independent optimization of 8 types of subsystems and output the single-iteration optimal solution of each subsystem; that is, the quantum particle swarm position update, orthogonal multi-population evolution, and Gaussian random walk perturbation are performed in parallel. S45. Heuristic Repair of Infeasible Subsystem Solutions: Based on the full-dimensional constraint set constructed in S3 and the single-iteration optimal solution of the subsystem output in S44, infeasible solutions with power balance and energy storage charge state exceeding the limit are identified, and the neighborhood time-series correction strategy is used to complete the repair of infeasible solutions, and the repaired feasible iterative solution of the subsystem is output. The power balance repair formula is as follows: ; The formula for correcting over-limit states of charge in energy storage is as follows: ; S46. Lagrange Multiplier Adaptive Update: Based on the feasible iterative solution of the subsystem output in S45 and the Lagrange multiplier vector initialized in S41, solve for the multi-energy flow power balance deviation. , , , , , Simultaneously, the Lagrange multipliers are adaptively updated using gradient descent. ; In the formula, and The first The second iteration and the first Lagrange multipliers in the next iteration; Update the step size coefficients for the multipliers; For the first Energy Flow Power balance deviation at any given time; S47. Determine if the following two-dimensional convergence criteria are met: The maximum number of iterations is reached here. ,and , If the threshold is set, the iteration is terminated and the optimal solution of the subsystem is output; otherwise, return to step S44.

[0020] Step S41 specifically includes the following steps: S411. Coupling Constraint Identification and Dual Decomposition Relaxation Range Delineation: Based on the grid interaction-driven two-layer-distributed hybrid optimization model of the multi-energy grid in the park, the Lagrange dual function, and the full-dimensional constraint set output by S3, identify and extract multi-energy flow power balance coupling constraints and cross-subsystem coupling constraints; clarify that the dual decomposition method only relaxes the cross-subsystem coupling constraints without changing the independent constraints and objective functions of each subsystem, and delineate the relaxation constraint range and retain the constraint boundary. S412, Multi-energy flow coupled constraint dual relaxation equivalent transformation: Based on the Lagrange duality principle, the original constrained optimization problem with coupling constraints in S3 is equivalently transformed into an unconstrained dual optimization problem. After relaxation, the objective function is consistent with the global dual function of S37. S413, Defining Independent Solution Boundaries for Distributed Subsystems: Based on the eight types of subsystems decoupled in S33, the independent solution boundaries, decision variable affiliations, and constraint memberships of each subsystem are defined. Specifically, the photovoltaic (PV) and wind power systems belong to the power supply side solution boundary, with decision variables including PV output, wind power output, and curtailment rate; the thermal, cooling, and hydrogen subsystems are independent energy flow solution boundaries, with decision variables including heating power, cooling power, and hydrogen production power; the energy storage subsystem is the cross-energy flow regulation solution boundary, with decision variables including charging / discharging power and state of charge; and the grid interaction subsystem is the external interface solution boundary, with decision variables including purchased and sold power. The output is a set of solution boundaries for the eight types of subsystems. S414. Distributed Solving Communication Rules and Iteration Parameter Initialization: Construct a master-slave distributed solving architecture. The master node is responsible for Lagrange multiplier updates, global convergence determination, and subsystem result coordination. The sub-nodes are responsible for independent parallel optimization of each subsystem. The sub-nodes only transmit power balance deviation information through the master node. Output distributed solving communication rules and iteration control parameter set. S415. Initialization of Full-Dimensional Lagrange Multiplier Vectors: Based on 6 types of balance constraints of the dual function in S37, for... Uniform non-negative initialization is performed, with all multipliers initialized to a constant of 0.1 to ensure balanced weights for coupling constraint penalties in the early stages of iteration; a complete initial Lagrange multiplier vector is constructed. ; S416. Validation of the distributed solution framework: The validation checks are performed sequentially on the equivalence of coupling constraint relaxation, the completeness of the subsystem solution boundary, the dimension matching of the Lagrange multiplier vectors, and the rationality of the iteration parameters. If the validation results are satisfactory, the output includes the coupling constraint relaxation results, the set of 8 types of subsystem solution boundaries, and the initial Lagrange multiplier vectors. Iterative convergence threshold If the verification fails, return to S412 and re-execute the coupling constraint relaxation operation.

[0021] Step S42 specifically includes the following steps: S421 and TG-OMPE-QPSO hybrid algorithm basic framework and subsystem adaptation parameter configuration: Based on the 8 types of subsystem solution boundaries, decision variable dimensions, and feasible region intervals of subsystem decision variables output by S41, configure the basic framework parameters of the TG-OMPE-QPSO hybrid algorithm to complete the matching and adaptation of the algorithm with the multi-energy flow subsystem; among which, the basic framework parameter configuration rules are: total population size Decision variable dimensions , For the first The decision variable vectors of each subsystem; the optimization time step is consistent with the dual-layer-distributed hybrid optimization model of the multi-energy network in the park; S422, Configure QPSO quantum behavior core operating parameters, orthogonal multi-population evolution core operating parameters, Tent chaos initialization core parameters, Gaussian random walk perturbation core parameters, convergence judgment and constraint processing parameters. Among them, the core operating parameters of QPSO quantum behavior include the contraction and expansion coefficients. Quantum attraction point weight Monte Carlo random factor Particle over-limit penalty coefficient ,and , No. The contraction and expansion coefficient of the next iteration The maximum number of iterations, , ; The core operating parameters of orthogonal multi-population evolution include orthogonal discrete levels. Multiple population ratios, ε-Box dominance threshold, orthogonal array type, and number of stalled iterations. ,and The population ratio is 70% for the basic population and 30% for the elite population. The ε-Box dominance threshold is used to optimize the return objective threshold. Minimum output target optimization threshold Orthogonal array type ; Tent's core parameters for chaos initialization include chaos control parameters. Chaotic sequence length Chaotic mapping interval ,and , The formula for generating a chaotic sequence in Tent is as follows: ; In the formula, and The first The and the first A chaotic sequence value; The core parameters of the Gaussian random walk perturbation include the perturbation mean. , disturbance standard deviation Perturbation probability Perturbation weights, and Take the mean of the decision variables of the globally optimal individual. , For the first The feasible region interval of the decision variable. , , The Gaussian random walk perturbation formula is as follows: ; In the formula, This represents the increment of the Gaussian walk perturbation; The numbers are Gaussian distributed random numbers. It is the globally optimal individual; This represents the current particle position. represents the constraint coefficient for the contraction of the feasible region boundary under Gaussian random walk perturbation. The adaptive perturbation weight adjustment coefficient for Gaussian random walk; Convergence criteria and constraint handling parameters include the algorithm convergence threshold. Subsystem solution accuracy threshold Infeasible solution repair threshold Constraint penalty function weights ,and , , ; S423, Algorithm Parameter Validation and Chaotic-Orthogonal Hybrid Initialization: Verify the dimensional matching, numerical rationality, and logical consistency of all configuration parameters; after passing the verification, execute Tent chaotic combined with orthogonal design hybrid initialization to generate the initial population; S4231. Generate a uniformly traversed chaotic sequence using the configured Tent chaotic initialization core parameters, and replace the orthogonally initialized random sampling points with the generated chaotic sequence. S4232. Using the orthogonal array type in the configured orthogonal multi-population evolution core operating parameters, the decision variables are... We perform symmetric discretization, and the discretization formula is as follows: ; In the formula, For the first 3D decision variables After being discretized, in the th Specific values ​​at each level; S4233. Map the random sampling points to the discrete intervals obtained in S4232 to generate the Tent-orthogonal hybrid initial population. ; S424, Elite Population and Basic Population Division: Based on the Hybrid Initial Population Generated in S423 Divide the basic population according to the ε-Box dominance rule. and elite groups This completes the final step of algorithm initialization.

[0022] Step S5 specifically includes the following steps: S51. Integration of Global Decision Variables: Based on the optimal subsystem solution output by S4, the time-series output data of each subsystem are uniformly organized according to the time-series dimension of typical scenario in S2 and the optimization time step in S3, and a global decision variable vector of the park's multi-energy network is constructed. ; S52. Verify whether the power grid peak-valley difference suppression rate is met. and and and and If yes, then the solution is deemed qualified and a candidate solution is output; otherwise, the solution is deemed unqualified and the process returns to step S42. in, ; ; ; ; In the formula: This refers to the peak-to-valley difference suppression rate of the power grid. To optimize the peak-to-valley difference; For grid power disturbance; For the first Energy Flow Constant global supply and demand imbalance; For the first Energy Flow Always supplying power; For the first Energy Flow Total load consumption at any given time; This is the allowable threshold for global balance deviation; This represents the total expected carbon emissions throughout the entire lifecycle of the park. Carbon emission allowances for the entire lifecycle of the park; S53. Multi-objective TOPSIS Global Collaborative Comprehensive Evaluation: Using economic optimization, power grid peak-shaving optimization, and low-carbon emission optimization as evaluation indicators, the TOPSIS approximation ideal solution ranking method is adopted to construct a comprehensive evaluation system for multiple sets of candidate global strategies, and the closeness of each candidate scheme is calculated: ; In the formula, For the first TOPSIS overall fit of the candidate solutions; For the first The Euclidean distance from the solution group to the ideal solution; For the first The Euclidean distance from the group of solutions to the negative ideal solution; S54. Determination and Standardized Output of Global Optimal Multi-Energy Mutual Aid Scheme: The candidate schemes are sorted in descending order of their proximity, and the scheme with the highest proximity is selected as the global optimal operation scheme for multi-energy mutual aid in the park under the power grid interaction.

[0023] Step S5 is followed by S6, which involves deploying IoT sensors to collect real-time operating data based on the optimal mutual assistance scheme output by S5. The scenario set of S2, the optimization model of S3, and the core parameters of the TG-OMPE-QPSO hybrid algorithm are dynamically updated through rolling time-domain optimization to form a closed-loop iterative optimization mechanism.

[0024] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for multi-energy mutual assistance in a multi-energy network within a campus, oriented towards grid interaction, characterized in that: Includes the following steps: S1. Construct a time-series feature library by integrating multi-dimensional data from the energy side, grid side, equipment side, load side, and interaction side. S2. Based on the time-series feature library output from S1, use Latin hypercube sampling-Cholsky decomposition-K-means clustering time-series coupled scenario generation and two-layer reduction methods to generate a set of typical multi-energy load scenarios and corresponding probabilities for grid interaction. S3. Based on the typical scenario set output from S2, construct a grid interaction-driven two-layer-distributed hybrid optimization model for the park's multi-energy network. S4. Based on the hybrid optimization model constructed in S3 and the TG-OMPE-QPSO hybrid algorithm, use dual decomposition to decouple the multi-energy flow subsystems and obtain the optimal solution for each subsystem. S5. Verify the optimal solution for each subsystem output from S4 and output the optimal mutual assistance scheme for the park's multi-energy network.

2. The method for multi-energy mutual assistance in a multi-energy grid in a park oriented towards power grid interaction as described in claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Simultaneously collect multi-dimensional data from the energy side, grid side, equipment side, load side, and interaction side; among which, the energy side data includes photovoltaic output. Wind power output Gas turbine output Hydrogen production capacity ; Grid-side data includes time-of-use pricing. Carbon trading prices Grid peak shaving depth Minimum inertia threshold ; Equipment-side data includes parameters for photovoltaic modules, wind turbine generators, gas turbines, combined heat and power (CHP) systems, electrochemical energy storage systems, thermal energy equipment, cooling energy equipment, and water electrolysis for hydrogen production and hydrogen coupling equipment. Photovoltaic module parameters include rated installed capacity. Photovoltaic module conversion efficiency Photovoltaic power temperature coefficient Irradiation intensity under standard test conditions With temperature upper and lower limits of photovoltaic power output constraints , Photovoltaic module floor area Operation and maintenance loss coefficient Photovoltaic curtailment power Photovoltaic operation and maintenance costs and the cost of solar curtailment penalties Wind turbine generator parameters include rated installed capacity. Fan cut-in wind speed Rated wind speed Cut-off wind speed Upper and lower limits of wind turbine output constraints , Wind energy utilization coefficient, wind turbine transmission efficiency The floor area of ​​the fan unit Wind turbine operation and maintenance cost coefficient Gas turbine parameters include rated power generation. Power generation efficiency upper and lower limits of gas turbine output constraints , Carbon emission factors throughout the entire life cycle Gas turbine operation and maintenance costs Cogeneration characteristic parameters include heat-to-power ratio. Heating efficiency Gas turbine floor space ; Electrochemical energy storage system parameters include rated capacity Rated charge and discharge power Charging efficiency Discharge efficiency Discharge power Charging power Self-discharge rate Upper and lower limits of charge state constraints , Depth of energy storage Cycle life mapping parameters, energy storage degradation cost coefficient Energy storage charging and discharging loss cost Energy storage operation and maintenance costs Energy storage unit floor area Charging power upper and lower limits , upper and lower limits of discharge power , ; Thermal energy equipment parameters include the rated heating power of the ground source heat pump. Coefficient of performance (COP) Output upper and lower limits and minimum operating load ratio, heating efficiency of gas-fired boilers Heating power of gas boiler Operation and maintenance costs of thermal energy equipment Thermal energy equipment floor area ; Parameters of cooling equipment include the rated cooling capacity of the ground source heat pump. Coefficient of performance (COD) Output upper and lower limits and minimum operating load ratio, refrigeration efficiency of electric refrigeration unit Electric cooling power Operation and maintenance costs of cold energy equipment Refrigeration equipment floor space ; Parameters of water electrolysis for hydrogen production and hydrogen energy coupling equipment include the rated hydrogen production power of the electrolyzer. Electro-hydrogen conversion efficiency Upper and lower limits of hydrogen production capacity constraints , Hydrogen energy storage rated pressure, capacity constraints and hydrogen energy utilization efficiency, and hydrogen production equipment operation and maintenance costs. and carbon emission factors Hydrogen storage and release power Hydrogen storage and charging power ; Load-side data includes electrical load. Heat load Cooling load Hydrogen load Time series data; Interactive data includes the power purchased by the park's power grid. Power sales volume of the park's power grid Total power exchanged between the power grid Grid interaction response commands, upper and lower limits of grid tie line power , ; S12. Perform data cleaning, neighborhood mean interpolation for missing values, and normalization on the original multidimensional data to obtain standardized multidimensional data. S13. Fit the probability distribution to the standardized multidimensional data to obtain the probability distribution model; S14. The Kendall rank correlation coefficient is used to calculate the time-series correlation of the source-load-storage multi-energy units of the multi-energy grid in the park, and the source-load-storage time-series correlation matrix is ​​obtained. S15. Integrate the probability distribution model and the source-load-storage time-series correlation matrix to construct a time-series feature library for the park's multi-energy network.

3. The method for multi-energy mutual assistance in a multi-energy grid in a park oriented towards power grid interaction as described in claim 2, characterized in that: Step S2 specifically includes the following steps: S21. Based on the time-series feature library output by S1, extract the park's features. Heat load Cooling load Hydrogen load Photovoltaic power output Wind power output Gas turbine output Power, hydrogen production capacity Total power grid interaction Nine sampling variables are used as sampling dimensions to obtain the sampling variable dimension set. The distribution function of each sampling variable was determined, the source-load-storage time-series correlation matrix was extracted, and a standardized time-series correlation coefficient matrix was constructed. ; S22. Latin hypercube sampling is used to achieve multivariate stratified unbiased sampling and generate independent initial scenarios. S221. Based on the sampling variable dimension set and empirical distribution function set output by S21, the probability distribution interval of each sampling variable is determined. Divided into equal parts There are three non-overlapping sub-intervals, each with a width of [value missing]. ; S222. Independently and randomly select uniformly distributed random numbers within each sub-interval, and calculate the sampling probability value of the sub-interval: ； In the formula, For the first The sampling probability value of each sub-interval; The numbers are uniformly distributed random numbers, and ; S223. Generate the first... through inverse transformation of empirical distribution. In the initial sampling scenario, the first The value of each variable : ； In the formula, For the first The sampling variable of the first One initial sampled scenario value; For the first The inverse function of the empirical distribution of each variable; S224. Arrange the initial scene values ​​of all variables according to the scene number and assemble them into a multi-energy load independent initial sampling scene matrix. , The total dimension of the source load random variables participating in the scenario sampling of the multi-energy network in the park. The total number of original, unreduced scenes generated by Latin hypercube sampling; S225. Determine the independent initial sampling scenario matrix. If the Euclidean distance between any two scenes is greater than a preset threshold, then the scene independence is deemed satisfactory, and the matrix of the independent initial sampled scenes is output. Otherwise, discard the sample and return to S221 to re-execute the sampling process; S23. Choleski decomposition to correct temporal coupling correlation: based on the independent initial sampling scene matrix output in step S22. The standardized time series correlation coefficient matrix output by S21 The temporal coupling relationship of the source load storage variables is restored and corrected to obtain the initial scene set of temporal coupling; S24. Initial Scenario Standardization and Power Grid Interaction Tag Matching: Based on the time-coupled initial scenario set output by S23, the scenario data is processed by maximum and minimum normalization. Then, combined with the power grid interaction time series parameters, the interaction constraint requirements are labeled for each scenario, and a standardized initial scenario set with power grid interaction tags is output. S25, First-level scene reduction: Based on the standardized initial scene set output by S24, the K-means clustering algorithm is used to reduce the dimensionality of the scenes and output the scene set with K-means cluster centers; S26, Second-layer scene reduction: Based on the K-means cluster center scene set output by S25, redundant scenes are eliminated using the probability distance criterion until the number of scenes is reduced to the preset number of typical scenes. The reduced candidate typical scene set and the initial probability value of the scene are output. S27. Scene probability redistribution and normalization: Based on the candidate typical scene set output by S26, the probability of the eliminated scene is redistributed and superimposed on the most similar retained scene. The probability of all retained scenes is normalized and corrected to ensure that the sum of the probabilities of all scenes is 1. S28. Scene Validity Verification and Error Assessment: Based on the typical scene set output in S27, combined with the original multi-dimensional data output in S1, the fitting accuracy of the scene is evaluated using the average fitting error as an indicator, and the temporal correlation retention of the scene is calculated. : ； In the formula, The temporal correlation coefficients for a typical set of scenarios; If the average fitting error of the scene And time-series correlation retention If the scenario is valid, the final set of typical multi-energy load scenarios, the corresponding scenario probabilities, and the validity verification report will be output.

4. The method for multi-energy mutual assistance in a multi-energy grid in a park oriented towards power grid interaction as described in claim 3, characterized in that: The hybrid optimization model described in step S3 has an upper layer of capacity configuration-grid interaction parameter optimization and a lower layer of multi-energy flow distributed operation optimization, embedding carbon cost, energy storage lifetime, and minimum inertia collaborative constraints.

5. A method for multi-energy mutual assistance in a multi-energy grid within a park, oriented towards grid interaction, as described in claim 4, is characterized in that: Step S3 specifically includes the following steps: S31. Construction of the upper-level optimization objective function: Based on the typical scene set and scene probability output by S2, construct the upper-level optimization three objective function; The first objective function is expressed as follows: ； In the formula, This represents the park's net income throughout its entire lifecycle. This represents the total number of typical scenarios output by S2. For the first The probability of a typical scenario; To optimize the total timing length; for Total carbon emissions of the park at any given time; for The cost of equipment maintenance at all times, and ; The expression for the second objective function is as follows: ； in, ； In the formula, This refers to the peak-valley difference in power exchange between power grids. The third objective function is as follows: ； In the formula, This represents the total carbon emissions of the system. Carbon emission factors from purchasing electricity from the power grid; S32. Add upper-level optimization constraints: Equipment capacity upper and lower limits constraints: Power constraints in grid interaction: ; Total investment and land use constraints: ; Constraints on curtailment of solar power: In the formula, For the first Installed capacity of this type of equipment; For the first Lower limit of capacity for this type of equipment; For the first Maximum capacity of this type of device; and These are the upper and lower limits for electricity purchases from the power grid; and These are the upper and lower limits for electricity sales from the power grid; This represents the total investment in equipment within the park. This is the upper limit for investment; The total area is, and ; This is the upper limit for land area; This is the upper limit of the light rejection rate; S33. Decoupling of Lower-Level Distributed Optimization Model and Construction of Subsystem Objectives: Based on the capacity configuration and grid interaction parameters output by the upper-level optimization, the dual decomposition method is used to decouple the global multi-energy flow optimization into eight types of distributed subsystems: wind power system, photovoltaic subsystem, park power subsystem, thermal subsystem, cold subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem. The objective function of the photovoltaic subsystem is as follows: ； In the formula, Operating costs of the photovoltaic subsystem; The objective function of the park's power subsystem is as follows: ； In the formula, The comprehensive objective function value of the park's power subsystem; for The cost of integrated operation, maintenance, and dispatch of the power grid within the industrial park at all times; The internal network power fluctuation penalty weighting coefficient; for Real-time fluctuations in aggregated power output within the park; for Equivalent inertia of the power subsystem in the park at any given moment; The real-time frequency change rate of the park's power subsystem; This represents the upper limit of the frequency change rate of the power subsystem. and These are the large penalty coefficients for exceeding the inertia limit and exceeding the frequency change rate limit, respectively. This is the function for determining whether a function exceeds its limit. The objective functions for the wind electronic system, thermal electronic system, cold electronic system, and hydrogen electronic system are as follows: ； In the formula, For the first The real-time operating cost of each subsystem, the first Each subsystem can be a wind electronic system, a thermal subsystem, a cold subsystem, or a hydrogen subsystem; For the first Cost function of each subsystem; For the first Subsystem Always put in the effort; The objective function of the energy storage subsystem is as follows: ； In the formula, Cost of the energy storage subsystem; The objective function of the power grid interaction subsystem is as follows: ； In the formula, For grid interaction costs; Output the distributed subsystem partitioning results and the single-objective operating cost function; S34. Add lower-level multi-energy flow power balance coupling constraints; Global power grid interaction and energy coupling constraints in the park: ; Thermal energy balance constraints: ; Power coupling constraints within the park's power subsystem: ; Cold energy balance constraints: ; Hydrogen energy balance constraints: ; Energy storage balance constraints: ; In the formula, It contributes to the integration of photovoltaic, wind power, and gas turbine industries in the industrial park; and They are respectively Time and The state of charge of the stored energy at any given moment; To optimize the time step; S35. Based on the interaction relationship of energy storage charging and discharging power in the power balance constraint of S34, construct the energy storage operation-lifetime collaborative constraint. Upper and lower limits of energy storage state of charge constraints: ； Energy storage cycle life constraints: ； In the formula, This represents the maximum number of energy storage cycles. S36. Based on the multi-energy flow power balance coupling constraint in step S34 and the energy storage operation-lifetime synergistic constraint in step S35, embed the grid interaction-low carbon-inertia synergistic constraint. System frequency change rate constraint: ; Carbon quota constraints: ; System equivalent inertia constraint: ; In the formula, The equivalent inertia of the power subsystem; for Carbon emission allowances for the industrial park at any time; S37. Based on the full-dimensional constraint set constructed in S35-S36, the Lagrange relaxation method is used to decouple all variables and construct a global dual function. : ； in, ； ； ； ； ； ； In the formula, This is a set of global decision variables, which includes photovoltaic power output. Wind power output Gas turbine output Hydrogen production capacity Energy storage charging and discharging power, total grid interconnection power , heating and cooling load power; These are the Lagrange multipliers for power balance, thermal balance, park power grid balance, cold energy balance, hydrogen energy balance, and energy storage balance, respectively. The total cost of the wind power system, photovoltaic subsystem, park power subsystem, thermal subsystem, cooling subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem; , , , , , These are the global power supply and demand balance deviation, thermal energy supply and demand balance deviation, power subsystem grid power supply and demand balance deviation, cold energy supply and demand balance deviation, hydrogen energy supply and demand balance deviation, and energy storage system energy balance deviation. This is the native electrical load within the park; For the interface power exchange between the park's power subsystem and the power grid interaction subsystem; Complete the model feasibility verification and output a complete dual-layer-distributed hybrid optimization model for the multi-energy network in the park.

6. A method for multi-energy mutual assistance in a multi-energy grid within a park, oriented towards grid interaction, as described in claim 5, is characterized in that: Step S4 Specifically, the following steps are included: S41. Dual Decomposition Method for Coupling Constraint Relaxation and Solution Framework Initialization: Based on the campus multi-energy grid two-layer-distributed hybrid optimization model, Lagrange dual function, and full-dimensional constraint set output from S3, the dual decomposition method is used to relax the global grid interaction energy coupling constraints, the campus power subsystem intragrid energy coupling constraints, and the multi-energy flow cross-subsystem coupling constraints of heat, cold, hydrogen, and energy storage. The distributed solution framework is initialized, defining the independent solution boundaries, inter-subsystem communication rules, and global iteration accuracy thresholds for the wind power system, photovoltaic subsystem, thermal subsystem, cold subsystem, hydrogen subsystem, energy storage subsystem, and grid interaction subsystem. The Lagrange multiplier vectors are initialized, and the coupling constraint relaxation results, subsystem solution boundary sets, and initial Lagrange multiplier vectors are output. , These are the initial multipliers for electrical energy balance, thermal energy balance, park power grid balance, cold energy balance, hydrogen energy balance, and energy storage balance. S42, TG-OMPE-QPSO Hybrid Algorithm Core Parameter Configuration and Population Initialization: Based on the subsystem boundary solution output from S41, the core operating parameters of the TG-OMPE-QPSO hybrid algorithm are configured; the algorithm population is initialized using a dual mechanism of Tent chaotic mapping followed by orthogonal design, resulting in the Tent-orthogonal hybrid initial population. ; S43. Encoding of decision variables and feasible region mapping for multi-energy flow subsystems: solving the subsystem boundary based on the output of S41 and the Tent-orthogonal mixed initial population based on the output of S42. The decision variables of the eight subsystems are encoded with real numbers; then, the encoded individuals of the population are mapped to the feasible domains of each subsystem, and matched with the full-dimensional constraint set constructed by S3 to obtain the feasible population of the subsystem; and the decision variable mapping formula is as follows: ； In the formula, For the first Initial population of each subsystem; For the first Each subsystem constraint set; S44. Parallel optimization of multiple subsystems: Based on the feasible population of subsystems output by S43, the TG-OMPE-QPSO hybrid algorithm is used to perform parallel independent optimization of 8 types of subsystems and output the single-iteration optimal solution of each subsystem. S45. Heuristic Repair of Infeasible Subsystem Solutions: Based on the full-dimensional constraint set constructed in S3 and the single-iteration optimal solution of the subsystem output in S44, infeasible solutions with power balance and energy storage charge state exceeding the limit are identified, and the neighborhood time-series correction strategy is used to complete the repair of infeasible solutions, and the repaired feasible iterative solution of the subsystem is output. The power balance repair formula is as follows: ； The formula for correcting over-limit states of charge in energy storage is as follows: ； S46. Lagrange Multiplier Adaptive Update: Based on the feasible iterative solution of the subsystem output in S45 and the Lagrange multiplier vector initialized in S41, solve for the multi-energy flow power balance deviation. , , , , , Simultaneously, the Lagrange multipliers are adaptively updated using gradient descent. ； In the formula, and The first The second iteration and the first Lagrange multipliers in the next iteration; Update the step size coefficients for the multipliers; For the first Energy Flow Power balance deviation at any given time; S47. Determine if the following two-dimensional convergence criteria are met: The maximum number of iterations is reached here. ,and , If the threshold is set, the iteration is terminated and the optimal solution of the subsystem is output; otherwise, return to step S44.

7. A method for multi-energy mutual assistance in a multi-energy grid within a park, oriented towards grid interaction, as described in claim 6, is characterized in that: Step S41 specifically includes the following steps: S411. Coupling Constraint Identification and Dual Decomposition Relaxation Range Delineation: Based on the grid interaction-driven two-layer-distributed hybrid optimization model of the multi-energy grid in the park, the Lagrange dual function, and the full-dimensional constraint set output by S3, identify and extract multi-energy flow power balance coupling constraints and cross-subsystem coupling constraints; clarify that the dual decomposition method only relaxes the cross-subsystem coupling constraints without changing the independent constraints and objective functions of each subsystem, and delineate the relaxation constraint range and retain the constraint boundary. S412, Multi-energy flow coupled constraint dual relaxation equivalent transformation: Based on the Lagrange duality principle, the original constrained optimization problem with coupling constraints in S3 is equivalently transformed into an unconstrained dual optimization problem. After relaxation, the objective function is consistent with the global dual function of S37. S413, Defining Independent Solution Boundaries for Distributed Subsystems: Based on the eight types of subsystems decoupled in S33, the independent solution boundaries, decision variable affiliations, and constraint memberships of each subsystem are defined. Specifically, the photovoltaic (PV) and wind power systems belong to the power supply side solution boundary, with decision variables including PV output, wind power output, and curtailment rate; the thermal, cooling, and hydrogen subsystems are independent energy flow solution boundaries, with decision variables including heating power, cooling power, and hydrogen production power; the energy storage subsystem is the cross-energy flow regulation solution boundary, with decision variables including charging / discharging power and state of charge; and the grid interaction subsystem is the external interface solution boundary, with decision variables including purchased and sold power. The output is a set of solution boundaries for the eight types of subsystems. S414. Distributed Solving Communication Rules and Iteration Parameter Initialization: Construct a master-slave distributed solving architecture. The master node is responsible for Lagrange multiplier updates, global convergence determination, and subsystem result coordination. The sub-nodes are responsible for independent parallel optimization of each subsystem. The sub-nodes only transmit power balance deviation information through the master node. Output distributed solving communication rules and iteration control parameter set. S415. Initialization of Full-Dimensional Lagrange Multiplier Vectors: Based on 6 types of balance constraints of the dual function in S37, for... Uniform non-negative initialization is performed, with all multipliers initialized to a constant of 0.1 to ensure balanced weights for coupling constraint penalties in the early stages of iteration; a complete initial Lagrange multiplier vector is constructed. ; S416. Validation of the distributed solution framework: The validation checks are performed sequentially on the equivalence of coupling constraint relaxation, the completeness of the subsystem solution boundary, the dimension matching of the Lagrange multiplier vectors, and the rationality of the iteration parameters. If the validation results are satisfactory, the output includes the coupling constraint relaxation results, the set of 8 types of subsystem solution boundaries, and the initial Lagrange multiplier vectors. Iterative convergence threshold If the verification fails, return to S412 and re-execute the coupling constraint relaxation operation.

8. A method for multi-energy mutual assistance in a multi-energy grid within a park, oriented towards grid interaction, as described in claim 7, is characterized in that: Step S42 specifically includes the following steps: S421 and TG-OMPE-QPSO hybrid algorithm basic framework and subsystem adaptation parameter configuration: Based on the 8 types of subsystem solution boundaries, decision variable dimensions, and feasible region intervals of subsystem decision variables output by S41, configure the basic framework parameters of the TG-OMPE-QPSO hybrid algorithm to complete the matching and adaptation of the algorithm with the multi-energy flow subsystem; among which, the basic framework parameter configuration rules are: total population size Decision variable dimensions , For the first The decision variable vectors of each subsystem; the optimization time step is consistent with the dual-layer-distributed hybrid optimization model of the multi-energy network in the park; S422, Configure QPSO quantum behavior core operating parameters, orthogonal multi-population evolution core operating parameters, Tent chaos initialization core parameters, Gaussian random walk perturbation core parameters, convergence judgment and constraint processing parameters. Among them, the core operating parameters of QPSO quantum behavior include the contraction and expansion coefficients. Quantum attraction point weight Monte Carlo random factor Particle over-limit penalty coefficient ,and , No. The contraction and expansion coefficient of the next iteration The maximum number of iterations, , ; The core operating parameters of orthogonal multi-population evolution include orthogonal discrete levels. Multiple population ratios, ε-Box dominance threshold, orthogonal array type, and number of stalled iterations. ,and The population ratio is 70% for the basic population and 30% for the elite population. The ε-Box dominance threshold is used to optimize the return objective threshold. Minimum output target optimization threshold Orthogonal array type ; Tent's core parameters for chaos initialization include chaos control parameters. Chaotic sequence length Chaotic mapping interval ,and , The formula for generating a chaotic sequence in Tent is as follows: ； In the formula, and The first The and the first A chaotic sequence value; The core parameters of the Gaussian random walk perturbation include the perturbation mean. , disturbance standard deviation Perturbation probability Perturbation weights, and Take the mean of the decision variables of the globally optimal individual. , For the first The feasible region interval of the decision variable. , , The Gaussian random walk perturbation formula is as follows: ； In the formula, This represents the increment of the Gaussian walk perturbation; The numbers are Gaussian distributed random numbers. It is the globally optimal individual; This represents the current particle position. represents the constraint coefficient for the contraction of the feasible region boundary under Gaussian random walk perturbation. The adaptive perturbation weight adjustment coefficient for Gaussian random walk; Convergence criteria and constraint handling parameters include the algorithm convergence threshold. Subsystem solution accuracy threshold Infeasible solution repair threshold Constraint penalty function weights ,and , , ; S423, Algorithm Parameter Validation and Chaotic-Orthogonal Hybrid Initialization: Verify the dimensional matching, numerical rationality, and logical consistency of all configuration parameters; after passing the verification, execute Tent chaotic combined with orthogonal design hybrid initialization to generate the initial population; S4231. Generate a uniformly traversed chaotic sequence using the configured Tent chaotic initialization core parameters, and replace the orthogonally initialized random sampling points with the generated chaotic sequence. S4232. Using the orthogonal array type in the configured orthogonal multi-population evolution core operating parameters, the decision variables are... We perform symmetric discretization, and the discretization formula is as follows: ； In the formula, For the first 3D decision variables After being discretized, in the th Specific values ​​at each level; S4233. Map the random sampling points to the discrete intervals obtained in S4232 to generate the Tent-orthogonal hybrid initial population. ; S424, Elite Population and Basic Population Division: Based on the Hybrid Initial Population Generated in S423 Divide the basic population according to the ε-Box dominance rule. and elite groups This completes the final step of algorithm initialization.

9. A method for multi-energy mutual assistance in a multi-energy grid within a park, oriented towards grid interaction, as described in claim 8, characterized in that: Step S5 specifically includes the following steps: S51. Integration of Global Decision Variables: Based on the optimal subsystem solution output by S4, the time-series output data of each subsystem are uniformly organized according to the time-series dimension of typical scenario in S2 and the optimization time step in S3, and a global decision variable vector of the park's multi-energy network is constructed. ; S52. Verify whether the power grid peak-valley difference suppression rate is met. and and and and If yes, then the solution is deemed qualified and a candidate solution is output; otherwise, the solution is deemed unqualified and the process returns to step S42. in, ； ； ； ； In the formula: This refers to the peak-to-valley difference suppression rate of the power grid. To optimize the peak-to-valley difference; For grid power disturbance; For the first Energy Flow Constant global supply and demand imbalance; For the first Energy Flow Always supplying power; For the first Energy Flow Total load consumption at any given time; This is the allowable threshold for global balance deviation; This represents the total expected carbon emissions throughout the entire lifecycle of the park. Carbon emission allowances for the entire lifecycle of the park; S53. Multi-objective TOPSIS Global Collaborative Comprehensive Evaluation: Using economic optimization, power grid peak-shaving optimization, and low-carbon emission optimization as evaluation indicators, the TOPSIS approximation ideal solution ranking method is adopted to construct a comprehensive evaluation system for multiple sets of candidate global strategies, and the closeness of each candidate scheme is calculated: ； In the formula, For the first TOPSIS overall fit of the candidate solutions; For the first The Euclidean distance from the solution group to the ideal solution; For the first The Euclidean distance from the group of solutions to the negative ideal solution; S54. Determination and Standardized Output of Global Optimal Multi-Energy Mutual Aid Scheme: The candidate schemes are sorted in descending order of their proximity, and the scheme with the highest proximity is selected as the global optimal operation scheme for multi-energy mutual aid in the park under the power grid interaction.

10. A method for multi-energy mutual assistance in a multi-energy grid within a park, oriented towards grid interaction, as described in claim 9, is characterized in that: Step S5 is followed by S6, which involves deploying IoT sensors to collect real-time operating data based on the optimal mutual assistance scheme output by S5. The scenario set of S2, the optimization model of S3, and the core parameters of the TG-OMPE-QPSO hybrid algorithm are dynamically updated through rolling time-domain optimization to form a closed-loop iterative optimization mechanism.

Images